Print

Mu, Di

Electrical Engineering

2007

MS

Rapid growth of the internet and multimedia communications has led to an unprecedented demand for broadband communication systems. Broadband communication performance relies on the characteristics of every element of the system. As the microelectronics steps into the deep sub-micron era, the VLSI operating speed and chip density have been tremendously increased. Unfortunately, these technology promotions do not only accelerate the operating speed, but also introduce an unwelcoming issue-timing jitter. Also, as the chip density increases, the dimension of the interconnect line shrinks, the characteristics of the interconnect line is becoming influential on the performance of the overall system. Hence developing an efficient model to study the timing jitter caused by the interconnect line is important for system design, synthesis and optimization. Data-dependent jitter (DDJ), one prominent category of timing jitter, is the deviation of each data threshold crossing time from a reference time due to the memory effect of previous data bits. Signals of different frequencies propagating over the interconnect line may suffer from different amplitude and phase distortion. The interconnect line has certain characteristics, i.e, limited bandwidth, which may induce the memory effect onto the signal transmission and decrease the signal integrity. This Master thesis characterizes the DDJ introduced by the interconnect line in the high-speed system. The research method is described as below: First, the transmission line transfer function is developed based on the ABCD matrices. Then, pattern extension method and Fourier analysis are adopted to develop the prediction model so as to sti~dy the input signals and channel responses. In addition, the interconnect line parasitics are extracted by adopting the third party CAD tool-FastHenry. Furthermore, the transfer function is simplified to first order system in order to minimize the DDJ. 'To validate our analytical model, our predicted results have been compared with the Matlab Simulink simulations. Results show that our proposed model can accomplish good accuracy.